-
This PDF is a selection from an out-of-print volume from the
National Bureauof Economic Research
Volume Title: The Regulated Economy: A Historical Approach to
PoliticalEconomy
Volume Author/Editor: Claudia Goldin and Gary D. Libecap,
editors
Volume Publisher: University of Chicago Press
Volume ISBN: 0-226-30110-9
Volume URL: http://www.nber.org/books/gold94-1
Conference Date: May 20-21, 1993
Publication Date: January 1994
Chapter Title: The Political Economy of Immigration Restriction
in the UnitedStates, 1890 to 1921
Chapter Author: Claudia Goldin
Chapter URL: http://www.nber.org/chapters/c6577
Chapter pages in book: (p. 223 - 258)
-
7 The Political Economy of Immigration Restriction in the United
States, 1890 to 1921 Claudia Goldin
It does not matter in the least what the favored classes of the
coun- try think about immigration; the doors of this land will
never be closed except upon the initiative and the imperative of
the laboring classes, looking to their own interests, and to the
heritage of their children.
Francis A. Walker, Discussions in Economics and Statistics
7.1 Introduction
With the passage of the Emergency Quota Act in May 1921 the era
of open immigration to the United States came to an abrupt end.'
The American policy of virtually unrestricted European immigration
was transformed, almost over- night, to a quota system that would
last, virtually unchanged, until 1965. The ultimate switch in
policy is not hard to explain. The perplexing part of the
legislative history of immigration restriction is its timing. More
astonishing than the closing of the door in 1921 is that it
remained open despite twenty- five years of assault during which 17
million immigrants from among the poor- est nations in Europe found
refuge in America. This paper details the remark- able set of
events that propped the door open and the forces that eventually
slammed it shut.
Because the story of immigration restriction is a legislative
one, its main players will be representatives, senators, and
presidents. But behind the legisla- tive tale are the shifting
interests of various groups. The first is organized labor,
Claudia Goldin is professor of economics at Harvard University,
director of the Development of the American Economy Program of the
National Bureau of Economic Research, and a research associate of
the National Bureau of Economic Research.
The author thanks Lisa Kao, Boris Simkovich, and Marian Valliant
for providing superb re- search assistance. Helpful comments were
provided by Stanley Engerman, Zadia Feliciano, Law- rence Katz,
Robert Margo, Jeffrey Williamson, the members of the Harvard
Economic History Workshop, and participants at the NBER-DAE
conference on the Political Economy of Regula- tion, particularly
the discussant, Joseph Ferrie. Shawn Kantor supplied the wage data
by city for the union sample, 1907 to 1923, and Howard Rosenthal
provided the congressional districts for the 1915 vote. The author
thanks them both. This research has been funded by National Science
Foundation grant SES-9122782.
I . As 1 will argue later, the abrupt end should more accurately
date with the final passage of the literacy test in 1917. since it
was a simple step to move from the test to a quota.
223
-
224 Claudia Goldin
represented by the American Federation of Labor (AFL) and the
Knights of Labor, and unorganized labor. Owners of capital, joining
together, for example, through the National Association of
Manufacturers, the National Board of Trade, and boards of trade and
chambers of commerce in numerous cities, are the second but the
most difficult to categorize. Immigrants, both new and old, are the
third.
There is also an important fourth group-rural America,
consisting of Yan- kee farmers as well as agriculturalists having
foreign roots. With one important exception, native-born rural
America was firmly in the anti-immigrant camp from the very
beginning of this story, and their anti-immigrant sentiment goes
back to earlier times2 This group was one of the major forces that
put the issue on the table in the 1890s, and they remained solidly
in the anti camp through- out. What shifts did occur in rural
America from 1890 to 1920 were a retreat from an open immigration
stance among older immigrant groups, such as Ger- mans and
Scandinavians in the upper midwestern areas, not a change of heart
among the native born.
The South provides the exception among rural native-born
Americans. Much of the South was in the pro-immigrant camp in the
1890s. But by the early 1900s the South had become a block solidly
against unrestricted immi- gration. I will have more to say about
this later.
Controlling segments of the various groups united in the 1890s
to form a coalition opposed to unrestricted immigration. The
coalition nearly succeeded in the late 1890s-indeed, they were but
two votes short of passing legislation to curtail immigration.’
Portions of the coalition switched sides during the first decade of
this century, and a new force to champion the cause of open immi-
gration-the recent arrivals themselves-emerged. Capital, which had
joined the anti-immigrant forces in the economically turbulent
I890s, threw much of its weight on the side of open immigration in
the early 1900s. Congress wit- nessed several battles over the
immigration issue during the twenty years fol- lowing the first
vote on the literacy test in 1897, but none succeeded in altering
the flow of immigration. It has been claimed that it took a world
war, igniting xenophobic and staunchly nativist sentiment, to pass
immigration restriction. There may be some truth to that view, but
the analysis in this paper suggests that the declining political
power of the foreign born, falling real wages for lower-skilled
workers after 1910, and the negative impact of the foreign born
2. See, for example, Higham L1955j 1981 on the two previous
waves of anti-immigrant and nativist sentiment in 1798, with the
Alien and Sedition Acts, and during the 1850s with the rise of the
Know-Nothing Party.
3. The closeness of both the I897 and 1898 votes belies the fact
that there was a large contingent in the House not voting on both
occasions, although some absences for the 1898 vote may be related
to Christmas recess. Further, about half those not voting in I898
were “paired,” that is, one nay and one yea who both agreed to be
absent (or two nay and one yea for the override vote). Although the
1898 vote did not clear a majority in the House, i t is termed
close because McKinley was president and would probably not have
vetoed the act. One cannot, however, rule out that such a large
number of abstentions may mean that the vote was not as close as it
seems.
-
225 The Political Economy of Immigration Restriction in the
U.S., 1890 to 1921
on the wages of even skilled workers may have eventually
clinched the vote for restriction.
The chronology of immigration restriction will be detailed
first. The history is well known and has been recounted elsewhere
(Higham 1955; Hutchinson 1981; Jones 1992; Taylor 1971). I then
move to a more in-depth analysis of city-level wage data by
occupation and industry from 1890 to 1923 to ascertain the possible
economic bases of support.
The wage data reveal substantial negative effects of immigration
for both laborers and artisans, although the effects by industry
depended on demand- side considerations. The impact, moreover,
appears to have increased from the 1890s to the early 1920s,
corresponding to the rise in negative sentiment toward open
immigration in the immediate pre-World War I period. Finally,
voting in the House is linked to the strength of the wage effect
and to the proportion of the population that was foreign born. The
greater the increase in wages in particular cities, the lower was
the probability that a representative would vote for restriction.
The greater the percentage foreign born in these cities, the lower
was sentiment for restriction. Once the foreign born had about a 30
percent share of a city’s population, support overwhelmingly
shifted to a pro-immigrant stance. At lower levels of the foreign
born-in the 10 to 30 percent range-the anti-immigrant position was
very strong, although at still lower levels it became weaker. The
desire to restrict was, therefore, tempered by the composition of
the electorate. Increased numbers of foreign born may have
threatened the economic position of many native-born workers, as
well as many foreign-born workers. The personal interests of the
foreign born in keeping the door open dominated economic interests
once the foreign born reached some critical level in a district.
But the foreign born may ultimately have been scapegoats for
unfavorable economic factors in certain local labor markets,
similar to recent experience in the United States.
7.2 The Literacy Test
7.2. I Chronology of Immigration-Restriction Legislation
The history of European immigration restriction in the United
States begins with the movement to pass the literacy test,
succeeding ultimately in 1917.4 Quotas and other types of blanket
restrictions were not seriously considered in the House or the
Senate prior to 1920.5 Of the multitude of regulations pro-
4. Immigration was restricted and regulated in various ways in
addition to the literacy test and, eventually, quotas, but none was
of great quantitative significance. Of most importance is that the
restrictions placed on Asians will not be treated in any detail
here. See, for example, Higham [1955] 1981 for a defense of
limiting attention to European immigration. It should be noted, as
well, that immigration from the Western Hemisphere was not
restricted by the 1921, 1924, and 1929 quotas, although the
literacy test was unaffected by that legislation.
5. Various influential groups, prior to the passage of the
quotas, had petitioned Congress to end immigration for some period
of time. The AFL in December 1918 requested that Congress
curtail
-
226 Claudia Goldin
posed, only two could have significantly restricted
immigration-the financial and literacy tests. Only the literacy
test received serious deliberation.h By the time the literacy test
finally passed, it was not as restrictive a measure as when it was
first proposed because literacy rose rapidly in Europe. Thus the
quotas of 1921, 1924, and 1929 quickly followed. The forces that
prompted these more restrictive measures were the same as those
that led to the passage of the literacy test. Thus most of this
paper is concerned with the passage of the literacy test, since the
quotas were its logical extension.
The literacy test was not merely given careful consideration in
Congress from 1897 to 1917. It passed the House on five separate
occasions and passed the Senate on four. Further, the House
overrode presidential vetoes of the bill twice and on two occasions
failed to override by fewer than seven votes. The Senate overrode a
presidential veto once, when the test became law in 1917.
The literacy test was to be administered to physically capable
adults to as- sess their ability to read. The test was
well-defined, although it varied some- what across proposed
immigration legislation. It generally consisted of reading several
sentences of the Constitution in any language chosen by the
potential immigrant, including recognized dialects. Some of the
proposed legislation also required that immigrants be capable of
writing the sentences they could read. Close relatives of an adult
male immigrant who was literate were often exempted. Because the
shipping companies that brought immigrants across the ocean were
responsible for the return voyage of any who did not meet U.S.
immigration standards, it is likely that these companies would have
adminis- tered a literacy test of their own, in the same way that
they screened for health violations in European p 0 m 7
immigration for at least two years (Higham [ 19551 1981). During
the debates over the quota Icgis- lation in the aftermath of World
War I, several bills were introduced that would have suspended
immigration for periods of from three to five years (Hutchinson 198
I , 17 I ) . Of the many possible means of restricting and
regulating immigration contained in the Reports of the Immigrarion
Com- mission of 1910, none was a blanket quota of the type
eventually adopted in I92 I, 1924, and 1929. One suggested means
would have limited “the number of each race arriving each year to a
certain percentage of the average of that race arriving during a
given period of years” (Senate 191 I a, 747).
6. Section 39 of the immigration bill introduced in 1906
contained a financial test that would have required, among other
things, that all male immigrants over sixteen years old (or the
male head of the household) have $25 or its equivalent (Hutchinson
1981, 139). The final version of the 1907 act did not contain the
provision. An amount of $25 was 2.4 weeks of income for lower-
skilled manufacturing labor in America in 1906 and about 9 weeks of
income for an equivalent worker in southern and eastern Europe at
the time (U.S. Bureau of the Census 1975, aeries D 778; Simkovich,
Taylor, and Williamson 1992).
7. The literacy test was put in place in 1917 and remained after
the quotas were passed. The experience with the literacy test
immediately following its passage, and prior to the quotas, can be
seen in U.S. Department of Labor 1918,23. The 1917 act allowed for
a fine of $200 per alien to be assessed against any transportation
company bringing an alien excludable by the literacy test. The fine
and the passage home may have been sufficiently steep to give
shipping companies an incentive to screen aliens prior to passage,
although I do not know whether or how they accom- plished that
task. In 1917 fines were levied for only 192 excludable illiterate
aliens out of a total of almost 300,000 aliens.
-
227 The Political Economy of Immigration Restriction in the
U.S., 1890 to 1921
The literacy test first came to a vote in Congress in 1897 and
was over- whelmingly passed by the House and cleared a majority in
the Senate (see the chronology in table 7.1). At least one other
bill was proposed during the debate in the House that could have
been even more restrictive and that would have restricted
immigration from any port in Europe not having a consular inspec-
tion station.
Several factors operated in the mid- 1890s to create a
short-lived coalition, yet one that would resurface in another
form, around regulating and restricting
Table 7.1 Immigration Restriction Chronology: Votes on the
Literacy Test
Date Branch of Government Vote Notes
2/9/97 2/17/97 3/2/97 3/3/97 3/3/97 111 7/98 1 21 14/98
512712
6/25/6
411 911 2
1211 811 2
21 1411 3 211 911 3 1/2/15 1/15/15
2/41 1 5 3/30/16 1211 411 6
2/1/17 2/5/17
House Senate President Cleveland House Senate Senate House
House
House
Senate
House
President Taft House Senate House
President Wilson House House Senate President Wilson House
Senate
217-36- 1 O P 34-3 1-25 Veto 195-37- 123
45-28-16 101 -1 04-150
No vote found
128-116
9-56-30
179-52
Veto 213-114-54
227-94- I03
Veto
307-87-39
Veto
50-7-39
261-136-26
64-7-25
287-106-40 62- 19-5
Affirmative vote on bill Affirmative vote on bill
Ovenides Presidential veto Takes no action, bill dies
Affirmative vote on bill Negative vote on
consideration of bill Affirmative vote on bill,
literacy test dropped in House-Senate conference
Vote to remove literacy test from immigration bill and to set up
Immigration Commission
Vote to strike the literacy test from the bill; affirmative vote
on bill, sent to conference
conference Affirmative vote on bill, sent to
Fails to override Affirmative vote on bill Affirmative vote on
conference
report of bill
Fails to override Affirmative vote on bill Affirmative vote on
bill
Overrides veto Overrides veto
Sources: Hutchinson 1981; Congressional Record, 62d, 63d, and
64th Cong. Note: Roll call votes count those not voting, whereas
non-roll call votes have only pro and con. aHutchinson reports
those not voting as 125, not 102.
-
228 Claudia Goldin
immigration. The leadership and members of the AFL and the
Knights of La- bor came out strongly in favor of the literacy test
in 1897, but had not done so before. The depression of the 1890s,
with its extremely high rates of unem- ployment, particularly in
the manufacturing sector, appears responsible for the change of
heart.* But capital, too, turned against immigration.
Industry had depended on immigrant labor. Thus the
restrictionist sentiment of certain associations of capitalists may
seem inexplicable. The labor unrest of the 1880s and early 1890s,
fresh in the minds of many, may have been a deciding factor. In
addition to a rash of strikes there were particularly odious
events, such as the Homestead Strike of 1892 and the Haymarket Riot
of 1886. The business faction that united against immigration in
the last two decades of the nineteenth century is not easily
categorized, but it disintegrated rapidly once economic conditions
improved, labor unrest subsided, and wage de- creases from
immigration were more apparent (Heald 1953; Wiebe 1962).
The face of immigration changed rapidly in the 1890s, moving
from north- ern and western Europe to southern, central, and
eastern Europe. Whereas the new immigrants were 35 percent of the
total flow in 1890, they were 56 percent in 1896, although the flow
was of comparatively modest size in the mid-1 890s, a product of
economic depression (see figures 7.1 and 7.2).9 Some have claimed
that the new immigrants were too recent and too few to motivate the
wave of anti-immigrant sentiment in the 1890s (Higham 1955). A
reading of the Congressional Record affords ample reason to
disagree with this claim, but not with a related assertion that the
new immigrants were too recent and too few to influence policy.“’
But they would be fortified by numbers and unified by fear very
soon.
President Cleveland vetoed the immigration legislation in 1897
because it contained the literacy test, and although the House
voted to override his veto, the Senate took no action and the bill
died. Just one year later, in 1898, a similar immigration law was
proposed in Congress. In this case the bill cleared the Senate but
failed by three votes to pass the House, which had just a year
before given it overwhelming support.lI The flip-flopping that took
place on this im-
8. The AFL letter to Congres, in I898 argued that “laborers are
imported from other countries to reduce our wagez and thereby our
standard of living” (Congressionnl Record 1898, 3 1686). The AFL,
like others, was arguing against contract labor and shipping and
railroad companies’ enticing people to emigrate to the United
States.
9. New immigrants are those from southern, central, and eastern
Europe. The countries (at various points in time) in the eastern,
central. and southern European group include Bulgaria, Croatia,
Czechoslovakia. Greece, Hungary, Italy, Montenegro, Poland,
Portugal, Rumania, Russia, Serbia, Spain, Turkey (in Europe),
Yugoslavia, and the Baltic republics. I have included non-
German-speaking emigrants from Austria in eastern Europe.
10. According to Higham ( [ 19551 1981) the Immigration
Protection League, organized primar- ily by the older immigrant
groups in the late 189Os, led the defeat of the I898 literacy
requirement in the House.
1 I . Of the 45 yeas in the Senate in 1898, 23 voted
altirmatively in 1897, 6 had voted negatively. 9 had been recorded
as absent. and 7 were new members of the Senate. Had all those
present in both I897 and I898 voted as they did in 1898, the vote
would have been 37 for and 22 againht in
-
Immigrants (Thousands) 1400
1200
1000
800
600
400
200
1880 1890 1900 1910 1920 1930
Fig. 7.1 Immigration, 1880-1930 Source: U S . Bureau of the
Census 1975, series C89-119. Note: New immigrants are those from
southern, central, and eastern Europe. See note 9 for the included
countries.
New Immigrants/Total Immigrants 0.8
0.6
0.4
1880 1890 1900 1910 1920 1930
Fig. 7.2 Proportion of new immigrants, 1880-1930 Source: U.S.
Bureau of the Census 1975, series C 89-119. Nore: New immigrants
are those from southern, central, and eastern Europe. See note 9
for the included countries.
-
230 Claudia Goldin
portant issue extended to the executive branch as well. Although
Cleveland vetoed the act in 1897, his successor, McKinley, ran on a
Republican platform that called for the literacy test. It is
doubtful that McKinley, having run on a platform calling for the
literacy test-the only such platform in the prerestric- tion
period-would have vetoed it. Had but two members of the House
changed sides in 1898, the literacy test would have become law,
although the large number of abstentions on the vote call the
apparent closeness into ques- tion. Thus the binding constraint in
1898 was the House, whereas the constraint just a year before was
the Senate.
The literacy test passed the House again in 1902, but was
dropped in House- Senate conference and was not again incorporated
into an immigration act until 1906. Through the political
maneuvering of Representative Joe Cannon, Speaker of the House, the
House voted in 1906 to remove the literacy test and set up the
Immigration Commission to explore the matter in greater depth. The
now-famous forty-two-volume Reports of the Immigration Commission
(U.S. Senate 191 lb) was issued in 1910. Seven methods to restrict
immigration were listed by the commission, including quotas, a
financial test, and a literacy test. “A majority of the
Commission,” the report concluded, “favor the reading and writing
test as the most feasible single method of restricting undesirable
immi- gration. . . . The Commission as a whole recommends
restriction as demanded by economic, moral, and social
considerations” (U.S. Senate 1911a, 1:48). On the heels of the
report, the literacy test was reintroduced in Congress in 1912.
From 1898, the previous vote on the literacy test in Congress,
to 1912, the next vote, were fourteen years of extraordinary
immigrant flow, particularly from southern, central, and eastern
Europe. The relative silence in Congress on the literacy test is
all the more curious. It might be claimed, however, that the halls
were actually not silent. There had been a vote in 1902, and the
test was almost incorporated into legislation in 1906. With the
creation of the Im- migration Commission, Congress may have felt
obliged to wait for its report, since its directive was to assess
immigration restriction. Another interpretation is that shifting
interests were at work. Although organized labor remained against
unrestricted immigration, capital had shifted decisively. Looking
more toward its long-run interests in holding down wages, capital
put aside its fears that labor unrest would be fueled by foreign
agitators. Perhaps of most impor- tance was the emergence of a
pivotal group in the form of the foreign born, who were vocal and
rapidly gaining the franchise.12
1897. The new members of the Senate in 1898 split their votes
about evenly for the test in 1898. Thus it was the disproportionate
exit of the negative votes, primarily Democrats, that increased the
strength of the prorestriction coalition, primarily Republican. See
Higham ([I9551 1981 1, who claims the bill passed the Senate in
1898 along party lines.
12. The new immigrants have been portrayed by many as a potent
force in big-city politics during the Progressive Era, but recent
data on the percentage foreign born in major cities who were
eligible to vote raises questions about their strength (see
Keyssar, forthcoming). Keyssar looks at the percentage naturalized
of males twenty-one years or older and finds that between 40
-
231 The Political Economy of Immigration Restriction in the
U.S., 1890 to 1921
Both the Senate and House passed the literacy test again in
1912, only to have it vetoed by lame duck President Taft. The House
failed by just six votes to override the veto. Had it been able to
override, the test would have become law, since the Senate vote was
86 percent in favor of the amendment on the literacy test. The
literacy test was reintroduced in 1915, passing the Senate by a
wide majority and the House by enough to override a veto but with a
large segment not voting. President Wilson, an ardent Progressive
remembering his promise to immigrants in the 1912 election, vetoed
the legislation, and the House failed to override it, this time by
just five votes.’7 In 1916 the House and Senate once again passed a
bill containing the literacy test, and finally, in 191 7, both
houses successfully overrode Wilson’s second veto. The literacy
test had become law.
The votes on the literacy test are evidence of the shifting
coalitions men- tioned earlier. The first vote in the House, in
1897, brought southern and urban- northeastern interests together
in opposition to the test, with virtually the rest of the country
favoring it.’4 The overwhelmingly anti-immigrant vote in 1897 may
have been a hysterical reaction to the prolonged economic downturn
of the 189Os, although recovery was well under way by the date of
the vote. A more sober view of the immigration issue may have been
given to the vote in 1898, a very close one in the House.
7.2.2 Analysis of Votes on the Literacy Test by State
As can be seen in table 7.2, New England, much of the Middle
Atlantic, and about half of the midwestern region were in favor of
the test in 1898. The South was generally against it, as it had
been in the previous vote. The Mountain and Pacific states were not
yet numerous enough to categorize. The next roll call vote on the
literacy test was not for another fifteen years, in 1913. By that
date the shifting coalitions mentioned earlier had become apparent.
The Northeast was split, with the larger cities voting
pro-immigrant and the rural areas voting anti. The Midwest was also
split. Differences there may have been rooted in
percent and 70 percent were, but that states often had residence
requirements that the mobile foreign born often could not meet. The
evidence presented here supports, in principle, the asser- tions of
the older literature. The foreign born might have been an even more
potent force had naturalization been faster and had various states
had more lenient residency requirements.
13. Wilson’s veto of the 1915 act can be rationalized, after the
fact, by his promise to the foreign born during the election, but
it is not clear that it could have been predicted prior to the vote
in the House. Only after the House passed the act did Wilson warn
the Senate that he would veto the bill if the literacy test was not
removed (Link 1954. 60-61). But there is no indication that Wilson
explicitly stated that he would veto the bill prior to its passage
in the House, although Link states that Wilson “intimated that he
would.” In fact, the official magazine of the International
Brother- hood of Teamsters predicted in August 1913, six months
after Taft’s veto of the literacy act, that “any immigration law
passed, carrying a literacy test in all probability, will be
approved by Presi- dent Wilson” (International Brotherhood of
Teamsters, August 191 3, 5).
14. Of the thirty-seven negative votes, twenty-five were cast by
southerners. Three from New York City joined them together with
eight others from urban areas in the Northeast. One additional
representative, from Wisconsin, voted against the test
(Congressional Record 1897, 29:2947).
-
232 Claudia Goldin
Table 7.2 Proportion of House Voting for the Literacy Test or to
Override a Presidential Veto of the Literacy Test, by State, 1898,
1913, 1915, 1917
Proportion to Override Number Voting For Test
- 1898 1913 1915 1917 1898 1913 1915 1917
New England CT 1 .0 MA 0.82 ME 1 .o NH 1 .0 RI 1 .0 VT 1 .0
Middle Atlantic NJ I .0 NY 0.52 PA 0.85
East North Central IL 0.47 IN 0.69 MI 0.50 OH 0.65 WI 0.43
West North Central IA 0.33 KS 0.0 MN 0.33 MO 0.13 ND 1 .0 NE
0.40 SD 0.50
DE 0.0 FL 0.0 GA 0.0 MD 1.0 NC 0.60 sc 0.57 VA 0.44 WV 0.67
East South Central AL 0.43 KY 0.22 MS 0.0 TN 0.22
West South Central AR 0.80 LA 0.0 OK - TX 0.17
South Atlantic
0.40 0.33 0.50 I .0 0.0 1 .0
0.0 0.28 0.65
0.39 0.80 0.44 0.74 0.40
0.27 0.86 0.44 0.79 1 .o 0.50 0.50
1 .0 I .O I .0 0.83 I .0 I .o 1 .o 1 .0
1 .o 0.91 0.89 1 .0
1 .0 0.17 I .0 0.80
0.0 0.25 0.67 0.0 0.0 1 .0
0.45 0.19 0.64
0.54 0.67 0.69 0.67 0.40
0.60 0.88 0.57 0.67 0.67 0.50 1 .0
I .0 1 .0 0.83 0.67 1 .0 0.86 1 .0 1 .o
0.8 0.91 1 .0 1 .0
1 .0 0.43 1 .0 0.78
0.20 0.35 0.80 1 .0 0.0 1 .0
0.67 0.26 0.75
0.63 0.50 0.50 0.80 0.55
0.80 0.86 1 .0 0.8 I 0.67 0.80 1 .0
1 .0 1 .0 I .o 0.83 1 0 0.86 I .o I .o
I .0 0.91 I .0 I .0
1 .0 0.63 1 .0 0.84
4 I 1 2 2 1 1
7 23 27
15 13 10 17 7
9 3 3
15 I 5 2
1 2
10 4 5 7 9 3
7 9 6 9
5 6
12
-
5 12 4 2 I 2
6 29 23
23 10 9
19 10
I I 7 9
14 I 6 2
1 1
10 6
10 6 9 2
9 I I 9
10
7 6 4
15
5 16 3 2 I 2
1 1 37 36
26 12 13 18 10
10 8 7
15 3 6 3
1 4
12 6 9 7
10 6
10 I I 8
10
8 7 7
18
5 17 5 2 3 2
12 42 36
27 12 14 20 11
10 7
1 1 16 3 5 3
I 4
12 6
10 7
10 6
10 I 1 8 9
7 8 6
19
-
233 The Political Economy of Immigration Restriction in the US.
, 1890 to 1921
Table 7.2 (continued)
Proportion to Override Number Voting For Test
1898 1913 1915 1917 1898 1913 1915 1917
Mountain Az CO ID MT NM NV UT WY
Pacific CA OR WA
Total
0.75 1 .0 0.0
I .0 1 .0 I .0 0.67 0.50 1.0 1 .o 1 .0 0.50 - 1.0 1 .0
0.0 0.0 1 .o 1 .0 1 .0 1 .0 0.0 0.50 0.0 I .o I .0 1 .0
0.71 0.91 0.90 1 .0 I .0 1 .0 I .0 1 .0 1 .0
4 1 I
28 I
1 3 1
2 1 1 1
-
I I I 10 I 3 3 3 5 4
342 409 427
Notes: A vote to override was a vote against open immigration.
“Paired’ votes (these were two to one for the override) are
included with either the yeas or nays. Those not voting (and not
paired) or absent are not included in the denominator. ‘Not yet a
state. hNo votes were cast by representatives of this state.
Source: Congressional Record, various years.
the nativity of constituencies, as they were in the cities.I5
The South was firmly against open immigration, as were the Pacific
region and most of the Mountain states. The 1915 and 1917 votes are
similar to that in 1913 with an erosion of support in much of the
Midwest and an increase in support in some large cities.
A large segment of rural America was against open immigration at
least by the first vote in 1897 and even in the first strongly
contested vote in 1898. Why this was the case probably has more to
do with the history of nativist sentiment in America than with the
particulars of immigration restriction of concern here. It is
important, however, that some parts of rural America were
considerably less in favor of restriction than were others. Rural
Minnesota, Wisconsin, Iowa, and Michigan can be easily contrasted
with equally rural areas in Ohio, Indi- ana, and Kansas (see table
7.2). In general, those from countries whose popula- tions were
still emigrating at high rates voted to keep the door open, while
the native born and those from countries that were not active
sending regions did
15. Ongoing research of mine on this issue indicates that those
of German and British ancestry opposed open immigration, whereas
those of Scandinavian and “new” ancestry supported it. I am also
exploring the role of concentration. Areas with many foreign born
of one nativity may have been pro-immigration. But areas with many
foreign born of several nativities may have been less willing to
keep the door open.
-
234 Claudia Goldin
not.I6 The reasons seem obvious, but one cannot differentiate
between a per- sonal interest in open immigration and an
ideological one. Recent immigrants may have wanted to send for
relatives and friends. But they may also have clung more fervently
to the notion that America was a haven for the world’s poor and
oppressed than did those who emigrated years before.
The proportion of the House vote in favor of the literacy test
has been re- gressed on the percentage foreign born, the level of
urbanization, and the im- migration rate from 1900 to 1910, all at
the state level. The regressions (table 7.3) demonstrate the
political strength of the foreign born but also reveal the mounting
opposition to immigration from residents of states with expanding
foreign-born populations.
The most obvious result in the regressions is that the constant
terms in all regressions are close to one-states with few foreign
born and only sparse urban areas voted overwhelmingly against open
immigration. Percentage for- eign born and percentage urban reduce
support for the literacy test. A one- point increase in the
percentage foreign born decreases support for the test by one to
two percentage points.” The greater importance of the percentage
urban variable in 1915 than in 1913 highlights the importance of
the redistricting that occurred between the 62d and 63d Congresses.
Although increasing the percentage foreign born reduces support for
the literacy test, an increase in the rate of immigration decreases
support.IX The impact of the rate of increase in immigration helps
explain the anti-immigration sentiment of the West.
The support the West gave to the literacy test by the 1913 and
1915 votes arose, it appears, from the rapid increase in the
percentage foreign born in those states, rather than from any
previous biases regarding Asian immigra- tion. With few exceptions
all states in the West had among the highest rates of immigration
but only moderate levels of foreign born. The percentage foreign
born in those states was insufficient to provide enough positive
sentiment against restriction, but the rate of increase was
sufficient to fuel a strong nega- tive reaction. Immigrants who
settled in the West during this period, it should be added, were
largely Europeans, not Asians and Mexicans.
The South has been omitted from the state-level regressions. Its
lack of for- eign born and paucity of cities would have lent
overwhelming support to the
16. A more accurate test of the proposition that the foreign
born from the current sending re- gions were more in favor of open
immigration than were those from regions that were no longer
sending a large fraction of their populations, requires
county-level data on nativity reaggregated to congressional
districts. I am currently collecting county-level data to
investigate the role of ethnicity and ethnic mix in immigration
restriction sentiment in rural districts and to explore how
statewide voting regulations affected the political strength of the
foreign-bom population. A sim- ple scanning of the ethnic origins
of populations in the midwestem states that were most antire-
strictive (e.g., Minnesota, Michigan, Wisconsin) suggests the
proposition stated here.
17. Adding a quadratic in the percentage foreign born shows that
the peak negative sentiment occurred at about 10 percent.
18. The immigration rate between 1910 and 1920 is used to gauge
the most recent flows of foreign born. Because immigration was very
low from 191 5 to 1920, most of the increase in the decade was
between 19 I0 and 19 15.
-
235 The Political Economy of Immigration Restriction in the
U.S., 1890 to 1921
Table 7.3 Explaining the Proportion of the House Vote for the
Literacy Test
Vote in 1913 = 0.857 - 2.08 (% FB) - 0.045 (% Urban) + 0.328
(Immigration Rate)
N = 31 (nonsouthern states); adjusted RZ = 0.44
Vote in 1915 = 1.02 - 1.19 (% FB) - 0.608 (% Urban) + 0.339
(Immigration Rate)
N = 32 (nonsouthern states); adjusted R2 = 0.63
(8.39) (3.31) (0.17) (2.50)
( 1 1.93) (2.30) (2.93) (3.37)
Nores: r-statistics, in absolute value, are in parentheses.
Ordinary least squares regressions are weighted by the total number
of representativeq voting. Vote in 1913 and in 1915 is the
proportion of the states’ votes in the House cast for the literacy
test (that is, to override the presidential veto); B FB is the
proportion of foreign born in the state in 1910; Immigration Rate
is the rate of increase in the foreign-born population between 1900
and 19 10. The vote in both years includes those who were “paired’;
in this case each negative vote was paired with two positive
votes.
relationships investigated. But such results would have offered
no insight why the South shifted sides from the 1890s to the early
1900s. The movement of the South from the pro-immigration camp to
the prorestriction side gave the prorestriction forces a decisive
edge in the House. Had the South remained in the pro-immigration
camp, the literacy test would not have cleared a majority in the
House even in 1915.” Several hypotheses for the switch can be
offered.
The most apparent hypothesis from a reading of the Congressional
Record is that antiforeign sentiment on the basis of race had
emerged. The South was struggling with its own race problem and
viewed the “new” immigrants as a European mulatto, adding yet
another racial group in America.*O Barring immi- grants from the
new sending regions would remove this danger and, moreover, would
not constrain the South. Southern states had tried to attract
immigrants, particularly for agricultural work, but they had not
flocked there in any num- bers for almost a century. The new
immigrants had gone north and west, rarely south.
Because the South had been unable to attract immigrants for some
time, its changed position around 1900 might have been related, as
well, to the resolu- tion of its own race problem. Jim Crow laws
may have given southerners the false sense that closing the door on
immigration would not lead blacks to flee en masse to the North.
Although their numbers might have been higher still
19. The total vote in 1915 (including the “pairs”) was 269 to
override and 140 against. The South accounted for 134 votes, and
118 were for the override, 16 opposing it. Had the South in 1915
voted as it had in 1898 (see table 7.2 for the proportions cast for
and against), it would have cast 43 votes for the override in 1915
and 91 against. The net gain for the pro-immigration forces would
have been 75 votes, giving the anti-immigration forces only 194 and
the pro-immigration forces 215.
20. See, for example, the speech of Senator Simmons of North
Carolina during the 1906 de- bates: “The broad fact, then, is that
about two-thirds of all the immigration to this country to-day and
during recent years has come from southern and eastern Europe.. . .
They belong . . . to a different civilization from that represented
by the Anglo-Saxon race” (Congressiond Record 1906.40: 7295).
-
236 Claudia Goldin
had blacks felt safe to leave, manufacturing interests in the
North did entice blacks to emigrate during World War I and
throughout the 1920s.
Southern manufacturing interests may have recognized that their
sole advan- tage was a low-wage, nonunion workforce, and that
immigrants were providing the North with a similar workforce. If
immigrants would not come South, the South would deprive the North
of them. Yet another potential explanation is that the North was
gaining power in Congress and that much of its population increase
was in the form of the foreign born and their children.21 Although
I cannot differentiate among these various hypotheses, each could
have been reinforcing. By the early 1900s the South saw nothing to
lose and much to gain from closing the door.
7.2.3 The Eventual Triumph of the Anti-immigration Forces
The three votes on the literacy test by three successive
seatings of the House enable one to see how the changed composition
of the electorate altered the outcome and precisely which forces
held the anti-immigrant forces at bay (see table 7.4).22 Comparing
first those representatives who voted in both the 62d ( 19 12/13)
and 63d (1 9 14/15) Congresses, 74 percent voted for the literacy
test. Thus the incumbent members of the House were overwhelmingly
in favor of restriction in 1915. The recently seated members of the
House did amass a majority in favor of restriction, but they did so
just barely. Only 54 percent voted for the test in 1915, clearly
not enough to override a presidential veto. Thus it was the newly
elected representatives who held the literacy test at bay,
suggesting that big-city districts had changed composition. The new
immi- grants themselves, it seems, managed to elect representatives
who voted dis- proportionately against the literacy test. But if
this were the only change in the House, the vote would have become
less in favor of the act over time. Rather, the percentage voting
in favor remained at 65 percent. Those who were voted out of office
were in favor of the keeping the door open to the same degree as
those who took their place. Thus the vote in 1913 would have
cleared the two- thirds needed to override, had only those who kept
their seats to 1915 voted. Those who were defeated in 1914 voted
far more decidedly against restriction, although with a majority in
favor of the literacy test.
Those who remained seated from the 62d to the 64th Congresses
voted dis- proportionately prorestriction in the 1913 and 1915
votes. Those newly elected and those who suffered defeat at the
polls in 1914 were less restrictionist. The
21. The South had opposed cheap land, a half century before, on
similar grounds. Cheap land meant more immigrants, and more
immigrants meant greater political power for the North. There were
additional reasons for southern reluctance to give land away. Cheap
land also meant higher tariffs, and the South opposed both high
tariffs and increased political power for the North. On the South’s
opposition to free land, see Robbins [1942] 1976.
22. I am looking only at the voting record of the House because
the Senate passed the test by wide enough margins in 1912/13,
1914/15, and 1916/17 to override a presidential veto. The Senate
would be expected to be more supportive of restrictive immigration
than the House, in which certain representatives were elected in
districts populated by the new immigrants.
-
237 The Political Economy of Immigration Restriction in the
U.S., 1890 to 1921
Table 7.4 Votes to Override the Presidential Veto on the
Literacy Test in the 62d, 63d, and 64th Congresses, 1913,1915, and
1917
Number For Number Against Number Not Voting % For“
Vote in 62d Congress (1913) Those remaining in office to 63d
Those defeated in 19 I4
Vote in 63d Congress (I 9 15) Incumbents only (62d and 63d)
Nonincumbents only (63d not 62d) Those defeated in 1916
Vote in 64th Congress (1917) Incumbents only (63d and 64th)
Nonincumbents only (64th not 63d)
213 I60 53
262 178 84 73
287 187 I00
I14 71 43
136 64 72 55
106 70 36
54 65.1 69.3 55.2
26 65.8 74.2 53.8 57.0
40 73.0 72.8 73.5
Sources: Congressional Record, various years; Congressional
Directory, various years
dTwo-thirds is necessary to override a presidential veto.
new members hailed primarily from the large and industrial
cities of the North- east and Midwest, whereas those defeated in
1914 came from small to middle- sized towns across America. Those
suffering defeat, therefore, were replaced by representatives far
less in favor of open immigration. But the newly elected group was
able to make up the difference and prop the door open. America had
become more bifurcated along the lines of open immigration, and it
was redistricting in 19 14 that resuscitated the pro-immigration
bloc.*’ Without it, the anti forces would have won. The increased
population of the nation’s big and industrial cities, with its
largely immigrant composition, was responsible for keeping the
anti-immigrant forces just below the two-thirds majority needed to
override. All that changed by 1917, however, when there was no
relationship between incumbency and the vote on the literacy test.
All in the House-save those whose districts were in the nation’s
largest cities and a handful of others-voted overwhelmingly for it,
regardless of time in office and party affiliation.
7.2.4 Restrictiveness of the Literacy Test
The literacy test was an overture to the Emergency Quota Act
passed in 1921, the Immigration Act of 1924, and, eventually, the
National Origins Act
23. The possibility that it was redistricting is by inference
only. There were forty-five more representative!, seated in the 63d
Congress than in the 62d Congress, and there were forty-three more
representatives present for the vote in the 63d than in the 62d
Congress to override the president’s veto (see table 7.1 ). Much of
the redistricting took place within states, it appears. A
tabulation of representatives by state does not reveal much
difference between the two Congresses. But New York City, for
example, gained seven representatives. Among those who were not
seated in the 62d Congress hut who voted in the 63d, there were
nine from New York City who voted against the test. Two
representatives from New York City were not reelected, one of whom
was against and one of whom was for the test. Three of the newly
elected representatives were from Philadelphia, which lost only one
seat from the 62d to the 63d Congresses. Chicago, however, made no
net gain.
-
238 Claudia Goldin
passed in 1929. Although the quotas were plausibly more potent
than the liter- acy test, the test could have imposed considerable
constraints, particularly on the newer immigrant groups. How much
of a constraint depended on the type of test, the sending country
flows, and the period considered.
As initially conceived in 1897, the literacy test involved
reading and writing a short passage of the US. Constitution and
barred illiterate adult males and their accompanying family
members. At that time it was believed that the test would have
checked the entry of 25 percent of all recent arrivals, although
more than 40 percent of the newer groups would have been barred.24
More precise estimates were compiled for the Reports of the
Immigration Commis- sion. According to the report, data collected
by the U S . commissioner general of immigration from the
self-reported statements of immigrants upon arrival indicated that
33.4 percent of eastern European and 44.9 percent of southern
European immigrants (fourteen years and older) arriving from 1899
to 1910 were illiterate.25 Thus the test would have reduced the new
immigrants by 37.4 percent in 1907 at the height of immigration.
The constraint would have been less in the 1920s due to the rising
literacy in eastern and southern Europe, although the test could
have been made more difficukzh
For the entire 1905 to 1914 period, a decade of immigrant flows
of more than one million per year, the literacy test would have
restricted immigration from southern and eastern Europe to about
445,000 annually when the flow was, in actuality, 7 1 2,000.27 But
the eventual quotas were far more restrictive. The 1921 act limited
southern and eastern Europeans to 156,000, and the 1924
24. During the debate on the immigration act of 1898, Senator
Fairbanks of Indiana inserted data in the Congressiunal Record
showing that about 25 percent of immigrants (fourteen years old and
over) arriving from I895 to I897 were illiterate. Illiteracy was
declared by the immigrant, and no official test was given
(Cungressiorzal Record 1898, 3 1 :5 15).
25. Female immigrants were less literate than male immigrants.
Because many versions of the literacy test allowed the illiterate
family members of a literate adult male immigrant to emigrate, the
constraint would have been less than calculated on the basis of the
aggregate data. But younger adults were more literate than older
adult immigrants, and since the Immigration Commission data group
all ages, this factor would tend to bias the calculation in the
other direction. The data from the U.S. commissioner general of
immigration in the Reports of rhe Immigration Commission ( 1 9 1 I
a, I :99) differ, often radically, by country from those reported
in the Congressional Record ( 1898, 3 1 :5 16) for a somewhat
earlier period of time. But the data in the report are consistent
with estimates I have computed using the 1910 Public Use Micro-data
Sample (PUMS).
26. Primary-school enrollment had been rising secularly in
Italy, Spain, Yugoslavia, and Ruma- nia across the latter half of
the nineteenth century and exploded in Russia after the revolution.
See, for example, the data in Easterlin 198 1.
27. Emigration to the United States from Europe could have
slowed in the 1920s as conditions improved in certain European
countries relative to those in the United States. Wage data
collected for a project on international economic convergence
(Simkovich, Taylor, and Williamson 1992) indicate that Italy, the
only new immigrant country in the data set, improved its real wage
position relative to the United States during the 1900s to 1920s
period. In 1910, for example, the ratio of Italian to American real
wages for unskilled laborers was 0.29, but by 1925 it was 0.48.
It should also be noted that even though gross immigration was
6.71 million from 1908 to 19 14, many immigrants returned home. The
net immigration figure is 61 percent of the gross, or 4.07 million
(Willcox 1931, 88).
-
239 The Political Economy of Immigration Restriction in the
U.S., 1890 to 1921
and 1929 acts lowered it further to just over 20,000, a mere
trickle. Put in terms of total immigration, from 1905 to 1914
730,000 would have entered each year had there been a literacy
test, whereas the 1921 act called for about half that number. The
1924 and 1929 acts stipulated numbers that were one-quarter to
one-fifth the hypothetical flows, Thus the literacy test, even as
conceived in 1897, would have imposed rather stringent restrictions
on the new immigrants, although not nearly as harsh as those
eventually imposed by the 1929 National Origins Act.
It is easier to understand why much of rural America lost
interest in immi- gration than it is to explain why it voted to
restrict immigration. Losing interest is not the same as feeling
threatened. The source of anti-immigrant sentiment could have been
nativism, anti-Catholicism, and racism. But another possibil- ity
is that many rural Americans, outside the South, saw the future of
their children, and possibly even that of their own, in the
nation’s cities and factories. It was in these cities that the
anti- and pro-immigrant forces waged their most contested battles,
and rural America may have sided with those who saw the American
standard of living threatened by immigrants. Urbanites, we shall
see, were pulled in two directions. The foreign born and their
children generally supported open immigration for the reunification
of their families and as a set of beliefs about America. But their
jobs and wages may have been threatened by unrestricted
immigration. The “heart strings’’ and the “purse strings” of urban
Americans often tugged in opposing directions. I turn now to the
eco- nomic effects of immigration to justify this characterization
of the support for anti-immigrant legislation.
7.3 The Economic Basis for Immigration Restriction
Almost all serious calls for the literacy test were preceded by
economic downturns, some of major proportion, and few economic
downturns of the era were not accompanied by a call for restriction
in the halls of Congress. Unemployment and labor unrest were
clearly in the minds of legislators in the 1897 and 1898 votes, and
economic conditions had worsened just as the 1915 literacy test
came to a vote. The major recession just following World War I was
a factor in the Emergency Quota Act. But the clamor for restriction
at particular junctures in our history must have been reinforced by
other eco- nomic forces, some national and long-run in nature and
some specific to the cities and periods that experienced the
greatest influxes. Immigrants, no matter where they went in the
United States, had economic effects on those already in the country
no matter where they lived and worked. But the initial impact that
immigrants had on wage levels of their close substitutes in
production must have been greatest in the local labor markets to
which the immigrants originally went and in which most remained.
The long-run story of general wage rate changes with the flood of
immigrants since the late 1840s is one of
-
240 Claudia Goldin
enormous importance on an international scale.’x That most
relevant to the political economy of restriction is a somewhat more
short-run tale.
The literacy test was introduced and gained momentum because
immigra- tion in the 1890s had shifted to ethnic and national
groups whose schooling levels and living standards were distinctly
below those of previous groups. They were, moreover,
disproportionately male and were often “birds of pas- sage” who
spent only brief durations in America. Such individuals were per-
ceived as a threat to the American working man. By toiling long
hours and bringing living standards from low-wage countries, they
probably did lower the wage-hours offer curve by more than an
equivalent increase in native-born workers would have. Moreover,
because they often lacked rudimentary skills in reading and
writing, and more often in speaking English, they may have earned
even less than competitive forces would have di~tated.’~ These were
certainly the claims of many observers of the day-Progressives,
conserva- tives, and labor movement organizers alike. Although each
group had its own solution, a dominant one was to restrict
immigration on the basis of literacy.
7.3.1 Occupations and Destinations of Immigrants, 1890 to
1920
Certain occupations and industries were disproportionately
composed of immigrants. If recently arrived immigrants were more
closely substitutable for other foreign-born workers and
lesser-skilled workers than for native-born higher-skilled workers,
then the wage effects should have been more negative in industries
and occupations having a large percentage of foreign-born and
lesser-skilled workers. The percentage of the labor force that was
foreign born by industry and for selected occupations in 1910 is
given in table 7.5. The foreign born are divided into three
groups-all foreign born; the “new” immi- grants, by which is meant
those from eastern, central, and southern Europe; and among the new
immigrants those who emigrated within the ten years pre- ceding the
19 10 census, termed “recent” immigrants.
All manufacturing employments were more heavily populated by
immi- grants than was the male labor force as a whole, although a
substantial fraction of the differential is accounted for by the
disproportionate employment of native-born workers in agriculture.
Excluding the agricultural sector, foreign- born workers were 1.4
times as likely to have been in the goods-producing sector than
were native-born workers, and the new immigrants were almost 1.6
times as likely.3” Among the industries most populated by the new
and recent immigrants were clothing, mining, and iron and steel.
But there was substantial variation in the ethnic backgrounds of
workers within industries; in foundries,
28. See Hatton and Williamson 1992 on the general issue of wage
rate changes with large-scale
29. See, for example, Hannon 1982 for empirical evidence on the
extent of labor market dis-
30. The goods-producing sector is mining, manufacturing, and
construction.
immigration on an international level.
crimination against immigrants during the late nineteenth
century.
-
241 The Political Economy of Immigration Restriction in the U S
. , 1890 to 1921
Table 7.5 Percentage of “New” and Recent Immigrant Males in the
Labor Force, by Industry and Selected Occupations in the
Goods-Producing Sector, 1910
( 3 ) (5) (1) ( 2 ) “New” (4) Relative %,
Born Immigrant Immigrant4 Foreign Born” and Recent” Foreign
“New” and Recent Relative %, “New”
(%) (%) ([1]/32.9) ([3]/12.2)
All employed males (2 14 years) Excluding those in agriculture
In goods-producing sector
Mining Building trades
Laborers Painters Brick and stone masons
Manufacturing Chemicals Clay, glass, and stone Clothing Food
Bakeries Iron and steel
Foundries Foundry laborers Machinists”
Leather Liquor and beverages Lumber and furniture Metals (except
iron and steel) Paper and pulp Printing and publishing Textiles
Tobacco and cigars
21.0 25.9 32.9 42.3 27.2 29.9 22.6 33. I 31.9 30.6 30.8 67. I
40.7 53.7 36.9 34.9 54.5 25.7 35.8 41.3 22.6 33.9 31.0 20.0 31.0
36.2
8.7 11.7 16.5 29.3 10.6 15.6 8.2
10.9 15.1 14.8 18.4 53.6 16.0 21.0 20.5 18.2 37.2 5.6
18.7 8.9 8.3
17.1 13.1 4.1
12.3 15.5
5.8 8. I
12.2 21.7 7.4
12.1 5.2 7.1
11.2 12.4 14.2 32.3 12.2 15.4 16.9 15.2 32.2
3.6 15.2 5.6 6.1
13.3 11.2 2.4 9.1 8.3
1 .OO 1.29 0.83 0.91 0.69 I .OO 0.97 0.93 0.94 2.04 I .24 I .63
1.12 1.06 1.66 0.78 I .09 1.26 0.69 1.03 0.94 0.61 0.94 1.10
1 .00 1.78 0.6 1 0.99 0.43 0.58 0.92 I .02 1.16 2.65 1 .OO I .26
1.39 I .25 2.64 0.30 1.25 0.46 0.50 1.09 0.92 0.20 0.75 0.68
Source: 1910 PUMS, males fourteen years and older. Note: “New”
and recent immigrants are eastern, central, and southern Europeans
who emigrated during the ten years preceding the 1910 census. “he
relative percentage divided by the percentage of all employed males
(fourteen years and older) in the goods-producing sector for each
of the two immigrant groups. hNot necessarily working in foundries
or in the iron and steel industry.
for example, 32 percent of the laborers were of the new and
recent group of immigrants but only 4 percent of the machinists
were.
Immigrants went disproportionately to the nation’s largest
cities, but so did all Americans during the period under study.
Despite the notion that immi- grants, particularly from 1900 to
1914, crowded themselves into a handful of America’s urban centers,
they were in fact extremely dispersed across all cities
-
242 Claudia Goldin
regardless of size.31 Indeed, the change in the foreign-born
population from 1900 to 1910 was, on average, the same across
almost all deciles of the size distribution of cities in 1900. The
fifteen cities with the largest and smallest increases in the
proportion of foreign born in their populations are given in part A
of table 7.6 for 1890 to 1900 and 1900 to 1910. No city in the top
decile (decile = 10) is included in the fifteen having the largest
increases from 1890 to 1900, and there are many small cities
represented among the ranks of those accumulating the foreign born
at a faster rate than they accumulated native- born residents. And
while there is some repetition in the top and bottom lists across
the decades, there is also a lot of movement. Immigrants went to
differ- ent cities in different decades. They went where the jobs
were, and, as will be demonstrated in table 7.7, they went where
their earning power would be highest.
Also of importance in assessing the political economy of
immigration re- striction is whether immigrants went to areas
already populated by immigrants. To the extent that “immigration
begot immigration,” certain cities and congres- sional districts
within them would have become even more disproportionately
immigrant in makeup and thus more inclined to oppose immigration
restric- tion. Part B of table 7.6 reports the results of the
regression of the difference in the percentage foreign born across
a decade on the percentage foreign born in the earlier year. That
is, A[% Foreign Born,, r + , O ) ] is run on [% Foreign Born,].
Interestingly, the coefficient is negative for the 1890 to 1900 and
1910 to 1920 decades, but positive for the 1900 to 1910 decade.’?
Immigration was reinforcing or concentrating in its impact from
1900 to 1910. Thus immigra- tion restriction was held at bay during
the largest immigrant flows, in part be- cause the new immigrants
were able to capture various congressional districts. By the 1910
to 1920 decade, however, the flows had a more diluting impact. Also
note that only during the decade of the greatest immigration, from
1900 to 1910, did immigrants flow into America’s cities at the same
rate that native- born Americans populated the same urban areas.
The percentage foreign born actually fell during the 1890 to 1900
and 1910 to 1920 decades in the cities under study. Similar notions
are apparent in part C of table 7.6. During the 1890 to 1900 and
1910 to 1920 decades, the percentage foreign born in the urban
population declined where population grew, but the reverse occurred
from 1900 to 1910. Only in the 1900 to 1910 decade did the
fastest-growing cities also increase their population share of the
foreign born. These bur- geoning urban areas gained representatives
who held the prorestriction move- ment at bay, at least for a
while.
31. The one exception-and it is an important one-is New York
City. There are 142 cities in the 1890 to 1900 sample and 127 in
the 1900 to 1910 sample. (These are the cities of the Bureau of
Labor Statistics wages and hours studies for the various time
periods.) The earlier sample in- cludes more small cities, although
the deciles in table 7.6 are recomputed for each decade.
32. The same cities have been used for the 1890-1900 and
1900-I910 regressions. There are twelve fewer cities for the
1910-1920 regression.
-
243 The Political Economy of Immigration Restriction in the
U.S., 1890 to 1921
Table 7.6 Changes in the Proportion of Foreign Born by City,
1890 to 1920
A. Chunges in proportion of foreign born in the populution
(AFB)a Largest Increases AFB Decileb Smallest Increases AFB
Decileb
I890 to I900 New Bedford, MA Passaic, NJ Hartford, CT
Bridgeport, CT Tampa, FL Middletown, CT Lincoln, NE Nashua, NH
Providence, RI Pueblo, CO Lynn, MA New London, CT Somerville, MA
Brockton, MA Schenectady, NY
1900 to 1910 Johnstown, PA Passaic, NJ Lynn, MA St. Joseph, MO
Brooklyn, NY' Utica, NY Trenton, NJ Elizabeth, NJ Youngstown, OH
Spokane, WA Bridgeport, CT Bayonne, NJ New Haven, CT Canton, OH New
Bedford. MA
,056 6 .055 2 ,025 7 ,023 7 ,018 I ,018 I ,014 4 ,013 2 .011 9
.010 2 .009 7 .008 I ,008 6 .006 4 .005 3
,072 3 ,056 3 .05 1 6 ,050 5 ,047 10 ,044 5 ,044 6 ,043 5 ,043 5
,043 7 ,042 7 ,042 3 ,042 8 .04 1 2 .039 6
St. Paul, MN Spokane, WA Duluth, MN Portland, OR Milwaukee, WI
Seattle, WA Davenport, IA Neenah, WI Tacoma, WA Saginaw, MI
Minneapolis, MN Holyoke, MA Chicago, IL Dubuque, IA Cincinnati,
OH
Davenport, IA Fall River, MA Covington, KY Clinton, IA Saginaw,
MI Fort Worth, TX Quincy, IL Troy, NY Oshkosh, WI Dubuque, IA
Evansville, IN Peoria, IL Salt Lake City, U'I Louisville, KY St.
Paul. MO
-.I12 -.I03 -.093 - .08 1 -.077 -.076
.07 1 - .070 - ,069 - .067 - .067 -.065 p.064 - ,063 - ,063
,052 -.050 - ,050 - .049 -.038
,037 - ,037 - ,036 -.035 -.033 -.031 - .028 - .027 p.026 -
,023
9 5 5 8
10 7 3 1 4 4 9 5
10 3
10
2 7 3 1 2 3 1 5 1 1 4 4 6 9 8
B. Regression of difference in % foreign born between I and ( t
i 10) on %foreign born in yeartd Dependent Variable Mean
% Foreign Born N R' Unweighted Weighted Coefficient (t-stat.)
on
1890 to 1900 -.I35 (-10.4) 127 .68 - ,0296 -.0373 1900 to 1910
,192 (1.86) 127 .27 ,0045 ,0131 1910 to 1920 -.I 19 (- 11.2) 115
.52 p.0298 .0390
C. Regression of difference in %foreign born between t und (t i
10) on log ofpopulation in year'd Coefficient (t-stat.) on
Log Population N R'
1890 to 1900 -.0041 (-3.20) 127 5 2 I900 to 1910 ,0053 (4.26)
127 .42 1910 to 1920 -.0057 (-5.31) 115 .I9
(continued)
-
244 Claudia Goldin
Table 7.6 (continued)
Sources: U S . Bureau of the Census, Cer7sus cfPopulution,
1890-1920. “he cities are those in the sample for the wage
regressions. There are 142 cities for 1890 to 1900 and 127 for 1900
to 1910. The change in the proportion of foreign born in the
population is calculated as (e.g., 1900 to 1910) percentage (white)
foreign born in 1910 - percentage (white) foreign born in 1900. It
is a percentage point change and is identical to the dependent
variable in the wage regressions in table 7.8. ”he city’s decile is
in the distribution of cities by population for 1900 and 1910. A
ten means the top decile, and a one is the lowest. cBrooklyn is
treated as a separate city in 1900 and 1910 for consistency with
the data for 1890, when it was independent. “All regressions are
weighted by the population in the base year. The 1890-1900 and
1900-1910 regressions also contain regional dummy variables; that
for I9 10-20 does not.
7.3.2 Wage Data by City, 1890 to 1923
Economists have, for some time, pondered the wage effects of the
enormous influx of less-skilled workers in the first two decades of
this century. Paul Douglas’s (1930) pioneering volume on wages from
1890 to 1926 concluded that real wages in manufacturing rose by 8
percent or only 0.32 percent average annually from 1890 to 19 14,
the period of greatest immigration. The increase from 1919 to 1926,
according to Douglas, was an astounding 3.3 percent aver- age
annually, whereas that in real wages in the several decades before
1890 was more on the order of 1.5 percent average annually.33 By
implication immi- gration had decreased the earning power of
manufacturing workers.
But Douglas’s findings were questioned by Albert Rees, whose
construction of a new consumer price index altered Douglas’s
central conclusion. According to Rees’s estimates, real wages rose
by 40 percent from 1890 to 1914, or 1.4 percent average annually (
[ 19611 1975, 120). By implication-and, once again, only by
implication since this is not a real test of the proposition-
immigration had not altered the course of real wages in the
manufacturing sector. The aggregate economy, it appeared, had
enormous absorptive capacity for new worker^.'^
But the data for the manufacturing sector (using Rees’s price
deflator), when contrasted with those for “lower-skilled” workers,
suggest that immigration depressed wages for the least skilled.
Figure 7.3 graphs the wage data for man- ufacturing workers and
those from Coombs ( 1926) for “lower-skilled” work- ers. Although
real manufacturing wages increased at about the same rate for the
entire 1900 to 1914 period, those for the “lower-skilled” workers
did not. The “lower-skilled’’ series slows down, flattens out, and
then declines some-
33. See Douglas 1930, whose series are reproduced in U.S. Bureau
of the Census 1975, series D 766 for nominal wages and E I85 for
the price index.
34. This is also a conclusion of Hatton and Williamson (1992)
based, in part, on Williamson (1982), who concludes, on the basis
of a computable general equilibrium model, that despite the
generally large absorptive capacity of the economy, it was lowest
around the World War I period.
-
245 The Political Economy of Immigration Restriction in the
U.S., 1890 to 1921
900
850
800
750
700
650
600
550
500
Real Annual Full-time Earnings (1 914 $)
- lower-skilled workers
- all in manufacturing
I I ---, I I I - . ,’
1900 1905 1910 1915 1920 1925
Fig. 7.3 Real annual full-time earnings, 1900-1924 Sources; U.S.
Bureau of the Census 1975, series D 740, D 778; Coombs 1926 for
1920-24. Nores: The weekly wages of the “lower-skilled” or Coombs
series is multiplied by 52 to get full- time earnings. Because the
lower-skilled data are defined as “full-time” on both a weekly and
an annual basis, they are inflated compared with those for all
manufacturing workers and are thus above the other series in two
years. Rees’s cost-of-living index ([ 19611 1975; see also U.S.
Bureau of the Census 1975, series E 186) for 1900 to 1914 is
spliced to the BLS Consumer Price Index for all items (series E
135); 1914 = 100.
time after 1907. Real wages among less-skilled workers stagnated
from 1900 to 1915.
Rees’s evidence, like Douglas’s, was indirect, and only by
inference could he conclude that immigrants had a slight impact on
the real wages of American manufacturing workers. He did not
directly estimate the effect of immigration on the wages of
workers. To get a more direct estimate of the economic impact of
immigrants would require a cross section of labor markets, each
receiving immigrants in different proportions to the existing
population. But a single cross section of cities may be
insufficient. Immigrants, as I will show, sought particular labor
markets that paid high wages. With city-level observations for two
cross sections one can estimate a difference equation that gets
around part of the simultaneity problem. Ironically, the same data
that both Douglas and Rees used to construct their nominal wage
series are precisely those that con- tain the type of observations
needed and used in this study.
Data on hourly and weekly wages for particular occupations and
industries by city are available for much of the period of
interest, although they are not uniform across the entire period.
For the 1890-1907 period there are the Bu-
-
246 Claudia Goldin
reau of Labor Statistics (BLS) wages and hours series for
nonunionized em- ployees that were used by Paul Douglas and Albert
Rees, among others. As many as one hundred cities were surveyed for
each of about twenty occupa- tions, with information on hourly
earnings given annually. For the 1907-23 period the BLS wages and
hours series covers unionized workers in thirteen occupations
across sixty-six cities.
In the data from I890 to 1907, two groups of occupations have
been selected for study. The first includes four types of
laborers-working in foundries, by contract on streets and sewers,
in municipal street and sewer work, and in the building trades, as
common laborers and as hod carriers. A second group in- cludes
skilled workers-painters, bricklayers, plasterers, plumbers, and
ma- chinists working in foundries and machine shops. The series
through 1903 is contained in the Nineteenth Annual Report (U.S.
Commission of Labor 1905) and is continued through 1907 in the
subsequent BLS wages and hours series, although with a reduced
number of cities. After 1907 the series covers only unionized
employees by occupation. In the data from 1907 to 1923 there are
only skilled workers and their helpers-bricklayers, carpenters,
wiremen and their helpers, painters, steamfitters and their
helpers, and iron finishers and their helpers. Both sets of
data-those for the nonunionized sample and the unionized-contain
hourly wages by year and occupation for a large number of cities.
That for the nonunionized group contains the number of workers in
the occupation-city cell, whereas that for the unionized group does
not.
Among the building tradesmen, laborers had about the same
proportion of new and recent immigrants as did the entire
goods-producing labor force. Painters and masons, however, were
disproportionately native born (see table 7.5), although a large
fraction of the masons were from older immigrants groups, such as
Germans. Among street and sewer workers 22 percent were the new and
recent immigrants, whereas only 12.2 percent of all in the entire
goods-producing sector were, yielding a relative proportion of 1
.8.35
City-level earnings data can also be found in the censuses of
manufacturing for 1899, 1904, 1909, and 1914. The data in this
source are by industry, not occupation. ,211 employees, not just
adult males, are covered, although for some of the industries men
were the bulk of the labor force. Annual earnings per production
workers, not hourly wages, are available for each of the four years
considered.
Four industries-men’s clothing, printing and publishing, bread
and bakery products, and foundries-have been chosen to span the
various characteristics of workers and products. The most serious
constraint on the choice of indus- tries was that the number of
cities represented had to be substantial, and not many industries
were found in a large enough sample of cities. Further, the
35. The data on street and sewer workers are not included in
table 7.5. Foreign-born workers were 49.4 percent of all street
laborers, the new immigrants were 30.5 percent, and the new and
recent immigrants wcre 22.0 percent.
-
247 The Political Economy of Immigration Restriction in the
U.S., 1890 to 1921
choice of industries was governed by the skills and ethnic
composition of workers. The nature of the product, as will be
apparent soon, was also a consid- eration.
Men’s clothing hired immigrant labor to a very large extent,
particularly tai- lors who came to America with training and who
worked in the production of coats that were traded nationally.
Printing and publishing, at the other end of the spectrum, hired
more highly educated laborers and very few immigrants- only 2
percent of its workforce were new and recent immigrants (table
7.5). The product was often locally consumed newspapers. Bread and
other bakery products, like men’s clothing, had large numbers of
immigrants among its workers and was found in virtually every city,
and like printing and publishing, its product was generally
nontraded. Foundries hired a mixture of skills and produced a
nationally traded good. Although foundry laborers were dispropor-
tionately new and recent immigrants, few machinists were.
The impact of immigrants on the wages of workers already in an
industry depends on the complementarity versus substitutability
between the two la- boring groups in the production function. It
also depends on how much immi- grants increase the demand for the
good produced by the industry. Immigrants increase the demand for
many types of goods, but their impact on local wages is greater and
more positive if these goods are produced locally. In terms of the
two main determinants of the impact of immigration on wages, the
four industries considered here can be categorized using the
following matrix:
Immigrants as a Percentage of the Labor Force Below Average
Above Average
_ _ _ _ _ _ _ ~
Product Demand Local Printing National Foundries
Bakeries Clothing
7.3.3 The Economic Impact of Immigration on Local Labor
Markets
The objective of this section is to estimate the impact of
immigration on the wage outcomes of native-born workers, in part to
assess whether immigration restriction was motivated by economic
concerns. Immigration to particular cities, like that to particular
countries, was not exogenous. Rather, immigrants went to cities
that had high wages. Thus a simple cross-sectional regression of
city-level wages on the percentage of immigrants yields a strong
positive coefficient, as is apparent in the regression coefficients
in table 7.7. But rather than indicating that immigrants caused
wages to increase, the result suggests that immigrants sought out
labor markets with high wages.
Certain cities could have had higher demand curves for
less-skilled labor than did others. If this higher demand were a
permanent feature of the city, as opposed to one that was
transitory, there is a simple way around simultaneity.
-
248 Claudia Goldin
Table 7.7 Cross-sectional Relationship between Immigrant Flows
and City Wages
A. Regression of hourly wages o n fraction of immigrants. by
city for various occupations, 1890-1 91 0
Using 1893 Wage, Using 1903 Wage, Occupations Elasticity" N
Elasticityd N
Laborers and hod carriers 0.094 192 0.135 192 Building trades
and machinists 0.101 278 0.082 278
B. Regression of annual earnings on frucrion of immigrants, by
city for various industries, 1900-191 0
Using 1904 Wage, Industries Elasticityd N
Bakeries Clothing Foundries Printing
0. I26 108 0. I25 48 0.078 101 0.092 105
Sources: By occupation: U.S. Commissioner of Labor 1905; by
industry: U.S. Bureau of the Cen- sus, Census of Manufactures,
1904; population: U.S. Bureau of the Census, Census of Populu-
rion, 1890-1920. Note; Fraction of immigrants = [foreign horn in (f
+ 10) - foreign born in r]/[average population from r to ( t +
lo)]. "he elasticities are evaluated at the means from a regression
of the wage in the year given on the percentage of the city
population that was immigrant, where immigrant = (foreign horn in
year t + 10) - (foreign born in year t ) . The regressions are
weighted by the number of workers in each occupation-city cell or
in each city-industry cell. When the 1893 wage is used, the
percentage immigrant is for 1890 to 1900; when the 1903 (or 1904)
wage is used, the percentage immigrant is for 1900 to 1910.
The method is to estimate a difference equation. The difference
in the (log of) wages for a group of workers is regressed on the
difference in the percentage of the population (or the labor force)
that is immigrant. The procedure, which estimates a fixed-effect
model, assumes that, for each city i, the (log) wage at time t ,
(wJ, is a function of the percentage foreign born, (TI), and an
error term consisting of a portion that may be correlated with F,,,
.sr or the fixed effect, and a portion that is not, (p,,,):
(1) Mw,,) = P" + P, (TI) + + F,,.
If equation (1) were estimated, the coefficient of interest, @,,
would be biased because cities that have positive demand shocks
will have both high wages and a high percentage foreign born. By
first differencing (and dropping the i subscripts) we get
-
249 The Political Economy of Immigration Restriction in the
U.S., 1890 to 1921
(see Altonji and Card 1991 for the functional form d e r i ~ a t
i o n ) . ~ ~ Note that fi,, which under the assumptions is now
unbiased, is the percentage change in the wage of a particular
group (e.g., artisans, laborers, workers in some industry) in
response to a percentage point change in the proportion of the
population (or labor force) that is foreign born.
Because immigrants can increase the demand for particular
products and thus the labor that produces them, as well as compete
with or complement other labor, the sign of p, is ambiguous a
priori. If the group in question were unskilled labor and if the
foreign born were disproportionately unskilled, then p, could be
less than or equal to zero. If the reference group were skilled
labor, however, p, could be positive.”
In addition to the potential biases already mentioned is the
possibility that labor, either native-born or prior immigrant
workers, migrates from cities in which recent immigrants landed.
This bias would result if recent immigrants drive away previous
workers by reducing wages, increasing the price of hous- ing, or
through a general dislike of the newer immigrant groups. Such
groups would then decrease wages in other cities that had fewer
immigrants. If mobile workers tend to equalize wages across cities,
the econometrician’s data would show little or no effect when there
was a negative effect for all workers of that skill “Spillovers” of
this type bias p, toward zero. The facts for the period under
consideration, however, do not suggest that native-born and al-
ready settled foreign-born workers were moving away from areas to
which recent immigrants went in the 1900 to 1910 period. Rather,
they were moving in. Although spillover effects could still bias
the relevant coefficient to zero, there is no evidence that the
bias was large.39
36. If the error term also consisted of a time-dependent
component, not orthogonal to F,, it would not be first differenced
away and could serve to bias the coefficient. Transitory demand
shocks would be such a factor and would serve to upwardly bias
p,.
37. There are two cases, one each for skilled and unskilled
labor. Altonji and Card (1991) present the unskilled case. The
skilled case is easily derived from their model and is given by
A log ws = {A,[(l - a)/(] - all + ~~ ,~A”(a /u )} /{ [q ,
”q~~(&” - IIJI - ( E , - qJ1 AUR where s = skilled labor, u =
unskilled labor, u = proportion of population that is unskilled, (Y
= proportion of immigrants who are unskilled, q = the usual
elasticities of substitution, E = the usual supply elasticities, I
= immigrants, P = population, and 0 5 A 5 1. The A’s are a function
of the degree to which the product is internally or externally
consumed. To the extent it is con- sumed by residents of the local
labor market, the positive impact of immigration on wages is
enhanced. Note that if q,, is positive, that is, the inputs are
relative substitutes, the effect of immi- grants on the wages of
the reference group must be 5 0. Only if the inputs are relative
comple- ments could the impact of immigration on the wages of the
reference group be positive. Because immigrants were
disproportionately unskilled, the impact of their increase on the
wages of the unskilled would have to be nonpositive. But there is
reason to view the skilled and unskilled as complementary, at least
in the short run. If the goods produced by the skilled (e.g.,
housing) are demanded by immigrants, the wages of the skilled could
rise with increased immigration.
38. The result will also hold if the effect were to increase
wages in occupations having workers complementary to
immigrants.
39. See, for example, Borjas, Freeman, and Katz 1992 on
estimating the economic impact of immigrants in a framework that
attempts to circumvent the spillover problem.
-
250 Claudia Goldin
The estimates of equation (2) are presented in table 7.8 for the
nonunion occupation sample (1 890-1907), the union occupation
sample (1907-23), and the industry sample (1899-1914). The data for
the percentage foreign born from the census are often, but not
always, for the nearest census date. In most cases the impact of
immigration is allowed to take effect over several years (e.g., the
equation for the difference in the wage from 1890 to 1903 uses
popu- lation data for 1890 and 1900).
The estimates of the impact of immigrants on the wages of
laborers are generally negative and often substantial, particularly
for the period extending into the twentieth century. Only the
artisan sample covers both the 1890-1907 and the 1907-23 periods,
and it shows an increase in magnitude of the effect with time. In
general, a 1 -percentage-point increase in the population share
that was foreign born decreased wages by about 1 to 1.5
percent.40
Interestingly, the negative effects of immigration on the wages
of both the unskilled and skilled occupations for the 1890-1903 (or
1907) period are not found for the 1890-97 period.4’ Wages appear
extremely rigid during the pe- riod of the 1890s depression and
only began to respond to the various labor market shocks with the
large change in prices after 1898. Thus when the liter- acy test
came before Congress for the first and second times (1 897 and
1898), capital may not yet have benefited from the wage effects of
immigration, but labor was still reeling from unemployment. By
1904, when capital had swung to the pro-immigration, antiliteracy
test camp, the wage effects were, in some cases, quite strong.
One may question the estimates showing substantial negative
effects of im- migrant workers on the wages of artisans. Immigrant
groups, particularly of the newer variety, were uncommon in several
of the artisan trades. Yet they may have been a threat to the wages
and employment of many artisan groups.42
40. The union sample uses weekly rather than hourly wages, while
the nonunion sample uses hourly wages. In the nonunion sample,
hourly wages are given, whereas the union sample has weekly wages
for a union contract and the contract hours for thc week.
Regressions using the implied hourly wage do not yield coefficients
that differ much from those using the weekly wage for the union
sample, but the standard errors are larger.
The estimates for the impact of immigration on wages are
approximately equal to those of Altonji and Card (1991) for the
less-educated native-born group. Comparisons between the two sets
of estimates, however, must be adjusted for slight differences in
variable definition. My csti- mates use foreign born as a
percentage of the total population, whereas they use foreign born
as a percentage of the labor force.
4 I . Note that the population data exist only for 1890 and
1900, but this is not the reason for the differences between the
1890-1903 regressions and those for 1890-97. The real reason is to
be found in the stickiness of wages, which may have been the single
most important factor giving rise to a large unemployment in the
1890s. A significant fraction of the cities had no change in
nominal wages from 1890 to 1897, but wages changed rapidly in the
face of price changes after 1897.
42. One possible bias concerns the convergence of wages across
cities. If immigrants went to cities with initially high wages but
wages were converging in the absence of immigration, a spuri- ous
relationship could be found between the wage change and the change
in the percentage forcign born. To test for this, I first checked
the data for wage convergence and next added the initial wage in
the equation. For 1890-1903 there was no wage convergence in the
sample cities. For 1907-23
-
251 The Political Economy of Immigration Restriction in the
U.S., 1890 to 1921
Table 7.8 Percentage Change in Wages with a Percentage Point
Change in the Proportion of Foreign Born: City-Level Observations
by Occupation or Industry, 1890 to 1914
P t-Statistics N
Laborersb 1890 to 1897 1890 to 1903 1890 to 1907
1890 to 1897 1890 to I903 1890 to I907
Artisans'
Artisans' I907 to 1915 1909 to 1915 1907 to 1923 1909 to
1923
By occupation, nonunion," hourly wage
-0.010 (-0.053) - 1.02 (-2.98) -1.60 (-3.39)
0.679 (2.92) -0.539 (-1.88) -0.145 (-0.33)
By occupution, union." weekly wage
-1.44 (-3.27) -1.20 (-3.58) -1.60 (-2.81) -1.41 (-2.65)
I92 I92 I60
278 278 162
223 223 225 225
By industqd annuul wage 1899 to 1914
Bread and bakery products 0.418 (0.69) 107 Clothing, men'sd
-3.06 (-2.45) 27 Foundry -0.829 (- 1.92) 91 Printing and publishing
0.764 ( I .47) 104
Sources: By occupation, nonunion: 1890-1903, U.S. Commissioner
of Labor 1905; 1907, Depart- ment of Commerce and Labor 1908. By
occupation, union: data provided by Shawn Kantor, from U S .
Department of Labor, 1907-23. By industry: U.S. Bureau of the
Census, Census ofManufac- tures, 1899-1914. Population: US. Bureau
of the Census, Census of Population, 1890-1920. Notes: Regressions
are estimated for each group of occupations or each industry. The
dependent variable is the difference in the log of wages between
the end and beginning years. Percentage foreign born is (foreign
born)/(total population). All regressions have been weighted by the
average number of sample workers in the interval, except those for
the union sample, where the weights are the log of city population
in 1910. The growth rate of the population (difference in the log
of the population between th